AI audio models on Kubeez

    Songs, instrumentals, four kinds of speech, stems and sound for video, each from the model built for that job.

    Try V6 and SoniloCompare models

    The lineup

    • Prompt: Bright modern pop. Punchy synth bass, tight claps, shimmering top line and a huge catchy hook about replaying the same memory. Confident female vocal, radio-ready polish.

      • V6
      • 194s

    • Prompt: Heroic trailer music. Massive brass theme, driving string ostinato, huge drums, choir-free, building from a quiet piano motif to a full orchestral climax.

      • V6
      • 138s

    • Prompt: Slow lo-fi beat, warm tape hiss, muted rhodes, no drums until the halfway point

      • Sonilo
      • 20s

    • Prompt: Slow smooth R&B. Velvet electric piano, muted trap-influenced drums, deep sub bass and a breathy layered female vocal. Midnight intimacy, lush harmonies, unhurried groove.

      • V6
      • 248s

    • Prompt: Dusty modern country. Acoustic guitar, lap steel, brushed drums and an honest male storyteller vocal about a quiet main street at closing time. Warm, nostalgic, wide open.

      • V6
      • 227s

    • Prompt: Hard-hitting boom-bap rap. Gritty sampled piano loop, heavy 808 sub, crisp snares and a confident male flow about building something from nothing. Cinematic street energy.

      • V6
      • 190s

    • Prompt: Anthemic alternative rock. Driving distorted guitars, live room drums, melodic bass, a raw male vocal and a huge soaring chorus. Late-night city energy, big and cathartic.

      • V6
      • 242s

    • Prompt: Modern latin pop. Nylon string guitar, reggaeton-leaning percussion, warm bass and a sultry Spanish vocal singing about the moon over the water. Romantic, sun-warmed, slow burn.

      • V6
      • 206s

    • Prompt: Melodic techno with a deep pulsing bassline, glassy arpeggios, airy processed vocal chops and a long hypnotic build into a euphoric drop. Dark, spacious, club-ready.

      • V6
      • 248s

    V6 and Sonilo

    Songs with vocals and your lyrics, or exact-length instrumentals.

    V6 is the default for natural vocals, V6 Wild for bold genre collisions, V6 Mini for fast sketches. Custom mode takes your lyrics, style and a male or female vocal. Sonilo never sings and returns exactly the length you ask.

    Best for

    • Jingles sung with your own lyrics
    • Full vocal tracks for reels and releases
    • Instrumentals cut to a fixed slot (Sonilo)

    Pick another model for

    • Scoring a specific clip: use Video Soundtrack.
    • Taking an existing song apart: use Music Separation.

    Specs

    Models
    V6, V6 Wild, V6 Mini, Sonilo
    Simple mode
    One prompt, 400 characters
    Custom mode, characters
    Title 80, style 1,000, lyrics 5,000
    Length, V6 family
    Optional, 10 s to 6 min
    Length, Sonilo
    Exact, 5 s to 6 min
    Try V6 and Sonilo
    See prices

    Price

    OptionRateUnit
    V634credits
    V6 Wild34credits
    V6 Mini34credits
    Sonilo0.75credits / sec
    Current rates in credits.

    Text To Dialogue V3

    One fixed voice reading approved copy word for word.

    Pick one of 26 voices, each with a preview, set the language and it reads exactly what is written. Bracket tags like [laughs] are removed, not performed. The API and MCP add stability, similarity, style and speed controls.

    Best for

    • Approved ad and product copy
    • Course narration in one consistent voice
    • Voiceovers in Romanian, Hindi or Arabic

    Pick another model for

    • Laughs, whispers or mood shifts: use a Gemini voice.
    • Two speakers in one file: use Gemini 3.8.

    Specs

    Voices
    26, each with a preview
    Languages
    29, including Turkish and Ukrainian
    Script
    5 to 5,000 characters
    Speakers
    One voice per render
    Try Text To Dialogue V3
    See prices

    Price

    OptionRateUnit
    Text To Dialogue V326credits
    Current rates in credits.

    • Prompt: Style instructions: Speak like a theatrical vampire host. Slow, velvety, menacing, savoring every word. Text: [whispering] Good evening. [long pause] I have been... [laughing softly] expecting you.

      • Algenib

    • Prompt: [laughing] I can't believe you actually did that. [laughing hard] That's the funniest thing I've ever seen.

      • Puck

    • Prompt: [whispering] Ve allí. En silencio, antes de que alguien te vea.

      • Kore

    Gemini 3.1 Flash TTS

    One voice you direct in plain words.

    Give it a script and a style prompt like "breathless sports commentator"; change the direction and the same words return as a new performance. Square-bracket cues like [sigh], [whispering] and [long pause] are acted, not read.

    Best for

    • Character lines that need real emotion
    • Hooks for Mexican versus Castilian Spanish
    • Languages V3 lacks, like Hungarian or Hebrew

    Pick another model for

    • Two speakers in one file: use Gemini 3.8.
    • Angle-bracket tags like <laugh>: only 3.8 performs them.

    Specs

    Voices
    30
    Languages
    87 language and regional codes, auto-detect
    Script
    4,000 bytes of UTF-8
    Style prompt
    4,000 bytes
    Speakers
    One voice per render
    Tags
    Square brackets, performed
    Try Gemini 3.1 Flash TTS
    See prices

    Price

    OptionRateUnit
    Gemini 3.1 Flash TTS26credits
    Current rates in credits.

    • Prompt: Okay, so we are live, and I have to ask. <short pause> Did you actually read the whole manual? / <laugh> Of course not. I read the first page, and then I just pressed every button until it worked. / That is the most honest answer we have ever had on this show. <chuckle> / <sigh> And honestly? It worked on the third try. <whispers> Do not tell the engineers.

      • Alnilam
      • Laomedeia
      • 24s
      • Kore
      • Enzo

    Gemini 3.8 Flash TTS

    A scripted conversation for one or two speakers.

    Write up to 50 lines, assign each to Speaker 1 or 2, and give any line its own delivery, like "dry, half asleep". Forty angle-bracket tags such as <laugh> and <gasp> are performed, not read aloud.

    Best for

    • Podcast-style exchanges between two hosts
    • Support, booking and phone-menu scripts
    • One narrator whose mood shifts by line

    Pick another model for

    • More than two voices in one file.
    • Square-bracket tags: this model does not perform them.

    Specs

    Speakers
    1 or 2
    Script
    5,000 characters, tags included
    Voices
    70, in 10 use-case groups
    Language
    Read from the text, no setting
    Output
    WAV, 24 kHz mono
    Try Gemini 3.8 Flash TTS
    See prices

    Price

    OptionRateUnit
    Gemini 3.8 Flash TTS26credits
    Current rates in credits.

    Seed 1.0 Audio

    A whole scene from one description: voices, effects, music.

    Write the scene like a director: place, music, sounds, who speaks and how, then the lines in quotes. Voices are described, not picked, or cloned from up to three clips tagged @Voice1 to @Voice3.

    Best for

    • Audio drama with music and effects
    • Game cutscenes and trailer moments
    • A character voiced from a real recording

    Pick another model for

    • Word-for-word approved copy: use Text To Dialogue V3.
    • Romanian, Arabic, Hindi or Ukrainian: outside its 20 languages.

    Specs

    Prompt
    Up to 2,048 characters
    Length
    Model decides, up to 120 s
    Voice references
    Three clips, 30 s each
    Image reference
    One image, not with voice clips
    Languages
    20
    API and MCP extras
    Format, speech rate, loudness, pitch
    Try Seed 1.0 Audio
    See prices

    Price

    OptionRateUnit
    Seed 1.0 Audio0.75credits / sec
    Current rates in credits.

    Music Separation

    Split a finished mix into vocals and an instrumental.

    Upload a mixed song and get two stems: the vocals, and everything else as one instrumental. Play, mute and download each in a waveform view with detected tempo. Dense masters can bleed, so listen before publishing.

    Best for

    • Karaoke and cover backing tracks
    • A cappella vocals for remixes

    Pick another model for

    • Separate drum, bass or guitar stems.
    • Isolating one singer from a duet or choir.

    Specs

    Input
    MP3, WAV, OGG, M4A, FLAC, AAC
    Output
    Vocals and instrumental, 320 kbps MP3
    Model
    BS Roformer
    Try Music Separation
    See prices

    Price

    OptionRateUnit
    Music Separation6credits / min
    Current rates in credits.

    • Prompt: Heavy wooden door slamming shut in an empty stone hall, long tail

      • 8s

    Sound Effects

    One described sound, rendered at the length you set.

    Write the sound, not the scene: what makes it, the space, how it decays. "Heavy wooden door slamming in an empty stone hall, long tail" works; "door sound" is a guess.

    Best for

    • Impacts, whooshes and title-card transitions
    • Room tone at the exact shot length
    • Interface and product sounds for demos

    Pick another model for

    • Effects synced to existing footage: use Video Soundtrack.
    • Music: use the music studio.

    Specs

    Input
    Text, up to 2,000 characters
    Length
    1 s to 3 min
    Output
    One audio file
    Engine
    Sonilo
    Try Sound Effects
    See prices

    Price

    OptionRateUnit
    Sound Effects0.6credits / sec
    Current rates in credits.

    Video Soundtrack

    Effects, a score or both, written against your footage.

    It watches the clip and writes effects matched to the action, a score paced to the cut, or both mixed in one pass. Get a bare audio track or the clip with sound in it.

    Best for

    • Silent AI renders and screen recordings
    • Trailers, ads and montages that build
    • Talking-head clips needing music under the voice

    Pick another model for

    • Very fast or ambiguous action: sound follows the picture.
    • Music without footage: use the music studio.

    Specs

    Input
    One video, up to 6 min
    Modes
    Effects, music, or both mixed
    Output
    Audio track, or clip with sound
    Notes
    Optional, up to 2,000 characters
    Engine
    Sonilo
    Try Video Soundtrack
    See prices

    Price

    OptionRateUnit
    Sound effects3credits / sec
    Music3credits / sec
    Music and effects6credits / sec
    Sound effects, with video3credits / sec
    Music, with video3credits / sec
    Music and effects, with video6credits / sec
    Current rates in credits.

    The eight audio tools side by side

    ModelYou give itYou get backLengthVoices
    V6 and SoniloPrompt or lyricsSong or instrumentalUp to 6 minMale, female or none
    Text To Dialogue V3A scriptOne voice, read exactlyFollows the script26 voices, 29 languages
    Gemini 3.1 Flash TTSScript and style promptOne acted voiceFollows the script30 voices, 87 codes
    Gemini 3.8 Flash TTSUp to 50 linesA conversation, WAVFollows the script70 voices
    Seed 1.0 AudioA scene descriptionVoices, effects and musicUp to 120 sDescribed or cloned
    Music SeparationA mixed songVocals and instrumentalSame as inputKeeps the original
    Sound EffectsOne sound, describedOne effectUp to 3 minNone
    Video SoundtrackA videoEffects, score or bothMatches the clipKeeps clip speech

    Which audio tool fits your job

    How the family fits together

    Audio is organised by job, not engine. The music studio pairs V6 songs with Sonilo's exact-length instrumentals. The Dialogue studio holds four speech models: a straight read, a directed voice, a two-speaker conversation and a full scene. Music Separation and the sound tools start from material you already have.

    Working tips

    • For a bed under narration, render the voiceover first, then ask Sonilo for exactly that many seconds.
    • On Video Soundtrack, run once with no note first: the model writes to what it sees.
    • For Seed 1.0 Audio, write cues as prose inside the scene, not as bracket tags.

    Pick a plan

    One credit balance covers images, video and audio. Pick the size that fits how much you make.

    Most popular

    Studio Pro

    Studio value with extra headroom.

    €49month
    6,400credits / month
    Team seats

    2 seats included

    What you'll generate

    • Images~1,600
    • Songs~457
    • Videos~640
    • Separations~1,066
    • Caption Minutes~213
    • All modelsIncluded
    Select plan

    Starter

    Low tier generations for regular use.

    €10month
    1,100credits / month

    What you'll generate

    • Images~275
    • Songs~78
    • Videos~110
    • Separations~183
    • Caption Minutes~36
    • All modelsIncluded
    Select plan

    Pro

    Access full experience.

    €20month
    2,600credits / month

    What you'll generate

    • Images~650
    • Songs~185
    • Videos~260
    • Separations~433
    • Caption Minutes~86
    • All modelsIncluded
    Select plan

    Powerhouse

    Most cost efficient.

    €99month
    13,000credits / month

    13,000 credits/mo

    Team seats
    3 seats

    Extra seats €19/seat/mo

    What you'll generate

    • Images~3,250
    • Songs~928
    • Videos~1,300
    • Separations~2,166
    • Caption Minutes~433
    • All modelsIncluded
    Select plan

    Enterprise

    For teams that outgrow standard plans.

    Custom

    Pricing tailored to your team

    What's included

    • Custom number of seats
    • Custom credit volume
    • Invoice billing
    • Priority support
    Contact sales
    Or pay-as-you-go

    Test any model with one-time credit packs

    Credits never expire and don't require a subscription — buy a pack, generate, see if Kubeez fits.

    Quick refill

    Quick refill for short bursts of generation.

    Credits

    500

    You pay

    €6

    Standard pack

    Mid-size pack at the standard per-credit rate.

    Credits

    1,500

    You pay

    €18

    −5%

    Volume pack

    Larger pack with a volume discount built in.

    Credits

    5,000

    You pay

    €60€57

    −10%

    Production pack

    Best per-credit price for sustained production.

    Credits

    15,000

    You pay

    €180€162

    Frequently asked questions

    • Yes. Make a Persona from a song you created, then pick it when you write the next one. A Persona sings your lyrics, so add lyrics when you use it.

    • Its limit counts bytes, not characters. Cyrillic, Greek, Arabic and Asian scripts use more than one byte per character, so fewer characters fit in one render.

    • No. Nothing is synthesised: the original voices stay, and the music can be ducked under them. To translate or re-voice speech, use Video dubbing.

    • Yes, through the API and MCP. Text To Dialogue V3 accepts the text before and after the line, so a short clip keeps the natural flow of the full script.

    Guides and comparisons

    Tests, comparisons and how-tos for Audio models from the Kubeez blog.

    Read more
    All blog posts

    Pick the right model and start

    Each model above links straight to the tool it runs in.

    Try V6 and SoniloAll model families